Resolving Grammatical Category Ambiguity in Acquisition

This kind of category ambiguity has been cited as evidence against the logical possibility of learning grammatical categories from distribution (Pinker 1987). Indeed, if we are to take seriously the many proposals that children may learn which words should be categorized together based on such cues as local co-occurrence (e.g., Redington, Chater & Finch 1998; Mintz 2003), we must address the potential pitfalls posed by such words. The difficulty with these words lies in the fact that their ambicategorical status may cause learners to conflate co-occurrence information across categories. That is, if a single word occurs with a distribution that is consistent with both noun and verb categories, learners could conclude that this distribution characterizes a single word class and thereby form a single category that contains all nouns and verbs. As Pinker (1987) pointed out, a child who hears the sentences in (3-5) would be expected to make errors such as (6).